Streaming, Multi-Screens and YouTube: The New (Unsustainable) Ways of Watching in the Home - Lancaster EPrints
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Streaming, Multi-Screens and YouTube: The New
(Unsustainable) Ways of Watching in the Home
Kelly Widdicks, Mike Hazas, Oliver Bates, and Adrian Friday
School of Computing and Communications, Lancaster University, Lancaster, UK
{k.v.widdicks,m.hazas,o.bates,a.friday}@lancaster.ac.uk
ABSTRACT to contribute to 21% of global electricity use by 2030 [2]. ICT
Internet use and online services underpin everyday life, and may exceed half the current relative contribution of green-
the resultant energy demand is almost entirely hidden, yet house gas emissions for the transportation sector by 2040 [8].
significant and growing: it is anticipated to reach 21% of Alongside this, Internet video takes a significant portion of
global electricity demand by 2030 and to eclipse half the our global online consumption (72% of consumer traffic in
greenhouse gas emissions of transportation by 2040. Driv- 2017) and has an expected Compound Annual Growth Rate
ing this growth, real-time video streaming (‘watching’) is (CAGR) of 34% for 2017–2022 [12, table 15]. Real-time en-
estimated at around 50% of all peak data traffic. Using a tertainment (e.g. on-demand video) contributed to 41.46% of
mixed-methods analysis of the use of 66 devices (e.g. smart fixed, and 32.95% of mobile, peak traffic demand for Europe
TVs, tablets) across 20 participants in 9 households, we reveal in 2015 [58]; these shares are even larger for North America,
the online activity of domestic watching and provide a de- taking 67.35% of fixed and 35.39% of mobile peak traffic in
tailed exploration of video-on-demand activities. We identify 2016 [59]. As network operators plan their capacity based on
new ways in which watching is transitioning in more rather peak traffic [57], growth in video streaming is a key driver
than less data demanding directions; and explore the role HCI in Internet infrastructure expansion and thus its climbing
may play in reducing this growing data demand. We further energy and carbon footprint.
highlight implications for key HCI and societal stakehold- There has been much discussion and dispute on the actual
ers (policy makers, service providers, network engineers) to environmental impacts of the Internet [2, 3, 8, 41, 61, 62, 70].
tackle this important issue. Yet, technology efficiency gains are quickly overtaken by
the rapid growth in Internet infrastructure [51] and service
KEYWORDS consumption [52], driving rebound effects [53] (‘Jevons para-
watching, video streaming, sustainability, data demand, de- dox’ [28]). Growing ICT energy footprint will likely continue
vices, households, everyday life, HCI, service design, policy to rise in tandem with increasing streaming popularity and
emerging technologies reliant on connectivity e.g. the Inter-
ACM Reference Format: net of Things (IoT) and Bitcoin [15].
Kelly Widdicks, Mike Hazas, Oliver Bates, and Adrian Friday. 2019.
We offer an in-depth analysis of month-long network
Streaming, Multi-Screens and YouTube: The New (Unsustainable)
Ways of Watching in the Home. In CHI Conference on Human
traces and qualitative interviews from 20 participants to un-
Factors in Computing Systems Proceedings (CHI 2019), May 4–9, pack the video-on-demand activities across 9 households. We
2019, Glasgow, Scotland UK. ACM, New York, NY, USA, 13 pages. identify the why, how and on what devices video streaming
https://doi.org/10.1145/3290605.3300696 is facilitated and how the activity of watching is changing
to incur more data demand (i.e. the “demand for network
1 INTRODUCTION connectivity and online services” [39, pg. 2729]). Given, our
How and when do we watch films, TV programmes or video community typically designs for more engagement which can
clips at home? We ask this at a time when digital devices and lead to data demand (e.g. by second screen apps [18, 24]), we
infrastructures (e.g. smartphones, data centres) are expected ask, what can be done to mitigate this energy impact and tran-
sition watching trajectories in sustainable directions [63]?
Permission to make digital or hard copies of part or all of this work for We propose implications for the HCI community and society
personal or classroom use is granted without fee provided that copies are in challenging the growth of video streaming.
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. Copyrights for third-
party components of this work must be honored. For all other uses, contact 2 RELATED WORK
the owner/author(s).
The practice of watching has been examined from a variety of
CHI 2019, May 4–9, 2019, Glasgow, Scotland UK
© 2019 Copyright held by the owner/author(s).
view points, and for many years. Such work has explored the
ACM ISBN 978-1-4503-5970-2/19/05. ways in which people watch TV [5, 60, 71]; analysed the adop-
https://doi.org/10.1145/3290605.3300696 tion, use and changes in specific services (e.g. YouTube [11],BBC iPlayer [34], Netflix [32]); investigated live streaming fixed-access networks for mobile consumption [33]. Other (e.g. on Twitch [64], in China [40], and viewers’ motivations work has focused on reducing the network load and en- for supporting streamers [73]); managing children’s screen ergy consumption of Internet watching, e.g. by predicting time [29]; discovered reasons for users multitasking their watched content and recording it via broadcast TV for an media [30, 31]; understood binge-watching [14]; the design ‘offline’ on-demand service [45]. Within sustainable HCI of second screen apps [18, 24, 44]; and “predictive preload- (SHCI), Preist et al. [51] have provided a rubric for address- ing” of video for work commutes [33]. Regrettably in terms ing digital infrastructure impacts (building on Blevis [9]). of sustainability, this prior work has not yet considered the They touch on designing for reduced data associated with energy used indirectly from watching related activity, and watching, e.g. by downloading episodes of a series a user may even add to the problem (e.g. with second screening). is watching at off-peak times or by creating ‘video on/off’ Mobile data traffic is growing at an alarming rate, increas- mechanisms for partially used streams [51]. More research ing 55% from 2016–2017 [21]; this growth is facilitated by is needed within HCI to discover what changes or interven- the roll-out and adoption of new infrastructure (e.g. 4G [49]) tions are feasible for users, or required in society to reduce and the opportunities for mobile data use they create (e.g. 5G watching data demand. and smart cities [27]). Thus, it’s no surprise that researchers Most recently, Rigby et al. [55, 56] have investigated on- in this domain have focused on the data demanding nature demand viewing across devices for 20 participants to un- of smartphones and tablets [39, 43, 72]. However, the bulk of cover contextual and situational factors of video streaming, global Internet traffic is still accessed through fixed networks including which and why different devices are chosen for (e.g. in-home Wi-Fi), demanding 67 Exabytes (EB) per month streaming, where it takes place, and who with. Whilst they (57 EB more than mobile) in 2017 with an estimated 225 EB highlight important findings (e.g. three quarters of watching by 2022 (a 27% CAGR 2017–2022) [12, table 15]. What’s more, sessions were carried out alone; YouTube is the most popu- there are many other types of Internet-connected (and non- lar service), their 14-day diary studies do not uncover data mobile) devices that have yet to be studied in sustainability demand, nor distinguish it across services and devices [56]. terms, including smart TVs and games consoles. In this paper, we scrutinise the activity and data demand To understand data demand from media and IT in the of watching in the home. We go beyond prior work by using home holistically, Bates et al. [6] suggest quantitative analy- a mixed methods analysis of qualitative interviews based on sis of network data. Network data has been previously logged analysis of month-long traces of network activity. This allows in the home, and video watching has been found to be one us to uncover the new ways in which watching data demand of the top Internet activities [35]. Yet the associated data is composed across different devices (including newer de- demand has not been discussed. As data demand “is designed vices such as TV dongles), services (including video watching into practices through its embedding in technologies and apps on Facebook), and types of users (and families). Our richly de- that we use to support them” [72, pg. 5361], we should explore tailed dataset provides the HCI community with insight into everyday life to discover what is causing such demand and video-on-demand activities. From this, we are in a unique how we as designers and engineers, can counteract it. position to highlight how HCI researchers and practitioners Streaming has been identified as a key category of traf- can adapt video streaming toward more sustainable Internet fic to target in HCI [72], as it augments peak electricity consumption, and discuss the broader implications for the demand [43]. Access to faster infrastructure has also been HCI community and society (e.g. policy makers) regarding positively correlated to increases in video-on-demand sub- Internet growth. scriptions [43]. Yet the energy associated with the Internet infrastructure is not simply correlated with active stream- 3 METHOD AND PARTICIPANTS ing; this relationship is more complex. Internet usage norms We gathered quantitative and qualitative data on Internet and follow a ‘Cornucopian paradigm’ [51]—i.e. growth in stream- device use in the home using a mixed-methods study, based ing leads to further growth in Internet infrastructure, and on established SHCI methods [6, 39, 72]. Twenty participants as a result, the total associated energy consumption also (nine households) took part (summarised in table 11 ) between increases. June 2017–January 2018, and were recruited through email Some designs have already been proposed to improve flyer advertisement and snow-balling methods. To protect users’ watching experiences that could possibly advance the the anonymity of the participants, pseudonyms are used in activity in more sustainable directions. These have included: this paper. removing video ‘auto-play’ to help address binge-watching behaviour [14]; providing short summaries of TV series to 1 Tim and Connie (H3) have two children (daughter aged 5, son aged 3), data prevent users re-watching entire episode collections prior demand was logged on their parent’s devices and is discussed in this paper, to new season releases [71]; and pre-downloading video on but they were too young to be interviewed.
H# Participant Pseudonym Personal Devices Shared Devices
(Age Range, Gender, Occupation) (Avg. Daily MB, No. of Days) (Avg. Daily MB, No. of Days)
H1 Ben (20s, M, Freelance Artist) Android Phone (911, 50), PC (476, 7), Amazon
Freeview Box (60, 57), Sonos Speakers,
Echo (21, 48), Kindle (11, 47)
Smart Meter
Gemma (50s, F, Retired) iPhone (63, 57), iPad (177, 56)
Martin (50s, M, Retired) iPhone (300, 57), iPad (195, 57)
Android Phone (31, 31), Amazon Fire Tablet (101, 33), Work Laptop (87, 16), Personal
H2 Laura (50s, F, Business Partner)
Laptop (29, 2)
H3 Tim (30s, M, Tax Consultant) Android Phone (1305, 34), Amazon Fire Tablet Smart TV (387, 34), YouView Box 1
(35, 3), Laptop (521, 7) (15, 36), YouView Box 2 (44, 36), An-
Connie (30s, F, Nurse) Android Phone (273, 33) droid Box (40, 1), Google Chromecast
(2899, 4), Windows Phone (19, 17)
H4 Alan (50s, M, Banker) Android Phone (3, 25), Windows Laptop (146, 9)
iPad (342, 29)
Denise (50s, F, Services Representative) iPhone (77, 29)
H5 Ella (30s, F, Lecturer) Android Phone (99, 27), Laptop (2716, 27)
Google Chromecast, Now TV Box
Kevin (30s, M, Researcher) iMac (216, 7), Android Phone (36, 1), Laptop
H6 Fred (40s, M, Accountancy Firm Employee) Android Phone (2435, 28), Work Laptop Amazon Fire Stick (453, 20), Laptop
Julie (40s, F, Medical Secretary) Android Phone (266, 28) (25, 16), Desktop PC (508, 9), Smart
Heather (10s, F, Secondary School Student) Android Phone (1696, 27), Android Tablet 1 (977, TV (47, 28)
20), Android Tablet 2 (655, 18)
H7 Ian (40s, M, Accountant) Work iPhone (157, 12), Personal iPhone (110, 10),
Sky Box 1 (704, 20), Sky Box 2 (1101,
iPad (66, 11), Laptop (171, 4)
32), Hudl Tablet (100, 16), Xbox 360
Olivia (40s, F, Community Coach) Personal Android Phone (36, 24), iPad (65, 9),
(170, 10), Printer (0.3, 7)
Laptop (285, 8), Work Android Phone
Nick (10s, M, Secondary School Student) Android Phone (318, 21)
Peter (10s, M, Primary School Student) Android Phone (0.2, 10)
H8 Rachel (30s, F, Accounts Assistant) iPhone (568, 27), Amazon Fire Stick (2634, 27), Laptop (373, 13)
Sally (0s, F, Primary School Student) iPod Touch (145, 26), Amazon Fire Stick (46, 27), iPad
iPhone (586, 23), iPad (2699, 25), Kindle (1, 23), MacBook Pro Laptop (99, 22), Sonos
H9 Xavier (20s, M, PhD Student)
(14, 25), TV (9, 19), PlayStation (5072, 21)
Table 1: A summary of the participants and their device use. The number of log days varies per device due to devices not
demanding data on every study day or logging issues for Ben’s PC and H7’s printer. Devices in italics were not logged in the
study (e.g. Olivia was unsure if work would permit her work phone being logged, H5’s shared devices were missed in setup).
To log Internet use, we replaced the home-router in each To understand data demand in the lives of our participants,
household2 and deployed a mini-PC to log and store Inter- the top 357 high level domain suffixes (responsible for 90%
net flows for one month duration (mean 35 days, max. 58, of total data demand5 from the 20,000+ domains found for
min. 26)3 . We logged the source and destination IP addresses all households) were manually mapped to services and then
and bytes transferred, and used logged DNS and DHCP re- categorised into activities. We take the view that it is not
quests to map each flow to human readable domains and particularly revealing to explore the ‘long tail’ 10% of least
participants’ devices. data demanding domains, which would also require manu-
For the qualitative data, we conducted two individual, ally mapping the remaining tens of thousands domain names.
semi-structured interviews with each participant: one before To provide an example of our mapping strategy (similar to
the logging phase to discuss their use of devices and the that used by previous work [72]), ‘pc-nowtv-ak.vod.sky.com’
Internet (mean duration 40 mins, max. 88, min. 16); and one mapped to the service ‘Now TV’ and was categorised into
after the logging phase to discuss visualisations of their quan- the ‘Watching’ activity. This process was carried out with
titative log data (mean duration 30 mins, max. 70, min. 12)4 . other non-watching domains, e.g. ‘i.instagram.com’ i.e. ‘In-
Each participant attended both of their interviews except stagram’ falls into ‘Social News and Networking’. Special case
Kevin (H5), who only undertook the first. Interviews were domains such as ‘video.xx.fbcdn.net’, where the service is
fully transcribed, open coded for themes, and then codes Facebook but the content is video, the category chosen is
were cross-compared for further analysis. ‘Watching’; domains associated with a watching device (e.g.
‘yv1-api.youview.tv’) are also categorised as this activity. If a
2 OpenWrt routers were used: https://openwrt.org/
3 Flows were captured using Cisco NetFlow: https://www.cisco.com/ 5 Only data demand which goes beyond the home network is explored in this
c/en/us/products/ios-nx-os-software/ios-netflow/ paper, i.e. the data transmitted between a device in the home and a network
4 Both interview schedules are provided as supplementary material. domain outside the home. The internal traffic (total: 3 GB) is omitted.domain outside the 90t h percentile of data demand had the H# Avg. No. of
Top Watching Services
same high-level domain name as a suffix, it was also included (Avg. Daily MB, No. of Days)
Daily MB Days
with the high-level domain as the same activity in order to H1 892 55
YouTube (803, 55), ITV Hub (184, 8),
Akamai (128, 8)
fully represent the demand of each watching service.
H2 5 31 Facebook Videos (7, 17), YouTube (0.8,
A total of 1,547 domains were classified as ‘Watching’;
26), Brightcove (2, 8)
these are the domains used within the analysis of this paper.
H3 1287 36 Warner Bros UltraViolet (1962, 6),
These values could potentially be under-represented due to: YouTube (310, 32), Watching Device
1) some domain services being difficult to decipher or their Unknown IPs (964, 10)
traffic/URL is ambiguous; and 2) participants’ freedom to H4 17 29 Brightcove (19, 10), YouTube (6, 28),
choose not to conduct certain activities during the study, BBC iPlayer (7, 11)
such as watching pornography. It is important to note that H5 2272 29 Now TV (3589, 13), Netflix (1658, 8),
watching data demand may not always be directly linked Facebook Videos (176, 27)
to immediate use by the end-user: due to background pro- H6 5145 28 YouTube (4882, 28), Facebook Videos
cesses (e.g. for a watching device), or a user not looking at (79, 28), Sky Sports (2120, 1)
the screen whilst it is demanding data (e.g. if the TV has H7 1690 32 Sky (1472, 32), YouTube (286, 24),
been left on). Determining this would require undesirably Facebook Videos (10, 5)
intrusive study methods such as video recording our par- H8 2749 27 TV Player (2480, 17), All 4 (747, 15),
ITV Hub (421, 14)
ticipants for the full study period. Furthermore, we clarify
H9 5738 24 YouTube (3370, 24), Netflix (2285, 16),
that the data discussed in this paper is associated with the
Twitch (910, 15)
home network and therefore is Wi-Fi only. We do not discuss
Table 2: Each household’s daily watching demand, the num-
mobile data, however this has been covered in-depth previ-
ber of days spent watching, and their top services.
ously [39, 43, 72]. We acknowledge that our sample size is
small in comparison with larger scale studies of network use,
but the purpose of our more detailed study is to uncover a ●
nuanced understanding of contemporary watching practices
indicative of wider trends in everyday life [12] and identify 750
how this links explicitly to data demand. ●
●
Data Demand (MB)
4 WATCHING DEMAND IN THE HOME 500 ●
● ●
All households (and 78% of devices shown in table 1) watched
●
● ●
some form of video content in the study (table 2). The services
●
used for this activity (e.g. YouTube, Netflix etc.) contributed
●
●
250
to 72% (558.57 GB) of the households’ total data demand ●
● ●
●
(777.09 GB). This activity typically occured every day of the
●
● ● ● ●
●
●
week (figure 1). There were peaks early morning (07:00) and 0
●
late at night (22:00), with an early evening peak at 18:00. 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3
Demand also occured in ‘prime time’ TV watching hours Hour of the Day
(20:00–22:00 [68]); this prime time watching is facilitated Figure 1: The average hourly data demand for watching
by devices other than smartphones, tablets and laptops, as for all households. The overlaid line and crosses represent
demand from portable devices dips at this time (figure 3)—a the mean; the dots represent the outliers. The figure is sup-
trough consistent with mobile devices [43, fig. 7]. ported by 9 months of fine-grained network logs. Some sam-
ples lie outside the inter-quartile range because data de-
Watching occurs across different device types (table 3).
mand is non-normally distributed.
Smartphones were the most commonly owned devices, 18 of
which accessed watching content, contributing 186 GB (33%
of total watching data demand). The most data-intensive on Facebook and Twitch. Domains associated with video
devices consist of H9’s games console (79 GB in the month- from Akamai also made this content delivery network a top
long study) and the TV dongles (avg. 20 GB per device). contributor. From this quantitative data we have uncovered:
Figure 2 shows the top ten services which compose the 1) when watching occurs; 2) the device types used; and 3) the
largest share of watching-related data demand. YouTube streaming services accessed. To understand how watching
was found to be the most data demanding service. Other related demand is formed; why, and how, this demanding
significant services consisted of Now TV, Netflix, Sky TV activity growing; and what parts of it are most meaningful
and TV Player, followed by more social-media related videos to users, we turn to our qualitative data.Device Type Avg. GB per device
Contributing way through, and I never would’ve done that like
(No. of Devices) (total devices’ GB)
Households a few years ago, I would’ve selected something,
Smartphone (18) 10.34 (186.17)
All households sat down and watched it, even if it was rubbish...
Tablet (12) 6.96 (83.63)
All but H5 there’s always something else that you could erm
Laptop (7) 9.4 (65.85)H2, H5, H6, H7,
be watching” (Ella, H5).
H8, H9
TV Box (5) 10.2 (51) H1, H3, H7 Ella and Kevin no longer have access to broadcast television
TV Dongle (4) 20.46 (81.84) H3, H6, H8 and describe it as “mundane”. Instead they pay monthly sub-
PC (4) 1.09 (4.37) H1, H3, H5, H6 scriptions to access the “massive amount of options” on three
Smart TV (3) 2.23 (6.71) H3, H6, H9 popular streaming services: Netflix, Now TV and Amazon
Games Console (1) 78.98 (78.98) H9 Prime. Whilst all other households had access to broadcast
Table 3: The different device types used for watching. television, some of them showed signs that follow the trajec-
tories of H5: Rachel and Sally (H8) only have access to con-
tent provided by Amazon Fire Sticks in their bedrooms; H9
YouTube 49.25%
has access to broadcast TV but “always stream[s]”; Gemma
Now TV 8.85% (H1) believes “you need the Internet for a telly”; H3 have ac-
Netflix 8.71% cess to some Internet-only BT channels; and Tim (H3) reflects
on whether broadcast television is even needed:
Watching Service
Sky 8.26%
TV Player 7.37%
“I’d still watch the same programs I watch I sup-
pose, yeah, I don’t like, I sit down to put the telly
Facebook Videos 2.51%
on and I don’t just watch whatever’s on at the time.
Twitch 2.39% If there’s nothing I wanna watch I just put some-
Akamai 2.12% thing on on-demand. I suppose it wouldn’t really
All 4 2.11%
matter if there was no broadcast telly.” (Tim, H3).
Warner Bros UltraViolet 2.06%
Furthermore, users default to streaming TV programmes or
0 100 200 300
films instead of accessing content through more traditional
Data Demand (GB) mediums. Alan (H4) tends to watch catch-up TV programmes
Figure 2: The total households’ data demand (GB) for the top rather than pre-recording them as it’s “just as easy” to do;
10 watching services and their % of watching demand. Other and in H6, Fred chooses to stream films rather than finding
watching services totalled to 35.5 GB (6.37%). his own DVD copy from his large collection at home.
YouTube: the most demanding watching service
5 THE NEW WAYS OF WATCHING
YouTube is used by all of the households and was found
To uncover why watching is such a large category of home to be the largest contributor to data demand—consuming
Internet use, this section explores the participants’ accounts, 49.25% (figure 2) of demand for watching across all house-
exposing new trajectories of watching-based data demand. holds (275.12 GB / 558.57 GB). For personal devices (i.e. de-
vices that are owned by one participant), the average daily
Streaming as the primary way of watching
total demand for YouTube was 648.25 MB for Generation Z
Video-on-demand provides a high-level of flexibility for watch- (participants born early 2000s onwards), 410.63 MB for the
ing activities: users can watch what they want, when they Millennials (1980s-2000s), and 186.02 MB for Generation X
want to, and on a multitude of devices at their choosing. Sub- (pre-1980s). This was evident in the interview discussions too
sequently, the nature of watching and what can be expected for Ben (H1) and Xavier (H9) (Millenials), alongside Heather
from this form of entertainment has changed, as Ella states (H6), Nick (H7) and Sally (H8) (Generation X).
below. Her household’s watching habits have become more While Heather (H6) knew the study was about device and
finicky since on-demand viewing (facilitated by their laptops, Internet use, she distinctly picked out YouTube to describe
Google Chromecast and Now TV Box) entered their lives: herself: “I’m 13, I play with my cats a lot, I go to school, and I
“Years ago you’d sit just for hours and watch Come spend quite a lot of time on YouTube”. Furthermore, Connie
Dine with Me or something on repeat on a Sunday, and Tim’s (H3) YouTube demand is extended by their 5 and
erm, but now we will start a series, if we don’t like 3 year-old children watching “ridiculous things where people
it after 1 or 2 episodes we will quit it and we’ll start are like playing with dolls” (Connie, H3).
watching something else, and we’re the same with For all generations, music was a particularly popular video
movies, like we constantly just turn movies off half type: Tim (H3) will watch music playlists with his family;Julie (H6) watched old music videos one night in the study; H# No. of Sessions per Day Durations of Sessions
and Sally (H8) accesses new music videos: “I let the music Mean Med Max Mean Med Max
play and I sing to it”. Yet, some participants do not always H1 0.1 0 2 2.3 2.2 4.4
watch the YouTube music videos they stream. Whilst Heather H3 10.1 8 54 112.3 32 853.4
H4 0.03 0 1 0.7 0.7 0.7
(H6) does a “bit of both” watching and listening, Nick (H7)
H5 1 0 6 6.9 4 41.5
will only listen to music videos via his phone whilst he plays
H6 12.5 11.5 23 38.4 9.9 441.6
on H7’s Xbox or researches online for school: “I don’t watch
H7 16 16 43 27 13.3 323.2
them I just put them aside to listen to music.” (Nick, H7). H8 3.6 3 9 3.9 2.6 17.1
Nick’s listening of YouTube videos for music began when H9 4.7 4 14 30.5 4 198.6
he got his first smartphone, but these habits can emerge in Table 4: The daily number and durations (mins) of multi-
other ways. H1’s free 3-month trial to Apple Music, listened watching sessions. H2 did not multi-watch. H3’s max dura-
to through their Sonos speaker, led Martin to develop a new tion means they streamed for a full day.
way of listening; this was continued after the trial via a new
Bluetooth speaker and streamed YouTube playlists due to
the paid-subscription nature of Sonos:
one device is streaming from any watching domain at a given
“The Sonos system, I’m a bit disappointed with”...“you time. Sessions are concatenated if they occur within 1 minute
can’t just stream things to it from Bluetooth, you’ve of each other, and must be at least 30 seconds in duration
gotta pay for a subscription to a music service”...“so (filtering out extremely short overlaps).
we’ve got a cheaper music speaker in there”...“I was For some, multi-watching is a rare or non-existent activity
using my phone and sending it via Bluetooth to (H1, H2, H4). However, it does happen at least once daily
that speaker, and it was, and it was fine, from, on for other households (H3, H6, H7, H8, H9). This is partly
YouTube, which is pretty good.” (Martin, H1). related to the number of people living in the home: H3, H6
Listening to music this way is much more data-intensive and H7 (the most frequent multi-watchers) are three of the
due to the video content involved. Whilst audio-only options largest households (three or four occupants each). Yet multi-
are available on YouTube Red/ Premium6 (a paid YouTube watching can also happen in a single-person household, as
service), Martin highlighted that they “don’t listen to [music] with H9 (table 4); Fred (H6) also discussed watching YouTube
frequently enough to pay a fee for it”. whilst watching TV–a data intensive activity via his Amazon
Fire Stick:
Multi-watching in the home
“I might be watching an episode or a film or some-
Watching separate content via different mediums at the same thing, if my phone’s there I might every now and
time (e.g. through broadcast TV, on-demand services, DVDs then go on and just look at Facebook or, I might
etc.) is a common activity for householders. We define this look at YouTube or something because I’ve short
everyday life reality as multi-watching, i.e. multiple, separate attention span of watching...” (Fred, H6).
watching activities occurring simultaneously in a given space.
Multi-watching via streaming is not yet routine. However,
This is enabled by devices (e.g. smartphones, tablets) existing
as we are increasingly turning to online content for watch-
in the home alongside the TV—for multi-watching to occur
ing, it is more than plausible to expect that multi-watching
previously, households would’ve had to own multiple TVs.
will be further accomplished by streaming in the future–
Multi-watching occurs in: different rooms of the home,
particularly as Internet speeds grow and allow for even more
e.g. Fred (H6) will stream Sky Go football on his study PC
simultaneous streams in higher qualities. (For example, the
when the lounge TV is in-use by his family; and in the same
UK Government aims for premises to have “full fibre” access
room, e.g. Fred described that Heather is “happy sitting on
i.e. fibre to the home by 2021 [26]).
the settee with us with her headphones in watching something
on YouTube”. Same room multi-watching has taken place in More devices, more watching demand
H3 too: Tim decided to stream BT Sport on his laptop since
The variety of devices available has created new possibilities
his children were watching the TV. Household members are
for how and when watching can be carried out. Portable
also streaming video for ‘solo entertainment’. H5’s Internet-
devices (i.e. mobile devices and laptops) are easily accessible
only watching allows Ella and Kevin to sit separately for this
and able to integrate “in areas of the home where comput-
activity–partially aided by Netflix’s profiles feature.
ing was previously unacceptable” [65, pg. 2642], allowing for
Table 4 shows each household’s daily number of streamed
watching activities to follow. “Communality and Portability”
multi-watching sessions; these occur when more than (any)
has been found to be the motivation behind mobile device
6 https://www.youtube.com/red video watching [35] and we extend this to laptops too; e.g.Portable Devices (37) Non-Portable Devices (17)
80
room, especially in winter, so we just go to bed.”
(Rachel, H8).
Data Demand (MB)
60
H3 arguably has the most complex watching setup of the
40 households. Alongside their mobile devices, they own a
Google Chromecast, an Android box, a smart LG TV and
20
two YouView boxes—all of which facilitate H3’s watching
0 in some way. Tim therefore has multiple options of how to
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3
watch the same TV programme or film, yet he chooses a
Hour of the Day
particular device configuration in order to watch in HD:
Figure 3: Average hourly watching MB by device portability. “...I get BT Sport, erm I only get standard defi-
nition on the YouView player, high definition on
the mobile app, so I tend to watch high definition
when prompting Ella (H5) about her large data demand on channels by connecting [his smartphone] to the
Wednesdays, she responded with: “you know what, Wednes- Google Chromecast and watching that on the telly
day is my cleaning day, that sounds ridiculous, but erm I watch, in high-def.” (Tim, H3).
I watch things while I’m tidying”. This was facilitated by her Throughout the study, Tim only ever accessed BT Sport
laptop, which she carries round the house to ‘wind down’ through his Google Chromecast and smartphone. These sce-
when cleaning or doing other chores. narios of how users go about watching sum up the complex-
Watching demand is greatest at 20:00–22:00 and is pre- ity of device setups—yet one thing is clear: they all lead to
dominately provided by non-portable devices (figure 3), yet more demand. Participants extend watching hours through
there are points in the day where watching on portable de- portable devices (and hence embed watching more into their
vices is more prominent: 07:00, 18:00, and 23:00. These times everyday lives); Ella will watch streamed content on larger
coincide with the mobile device watching demand peaks [72, devices; Fred will stream football matches instead of watch-
fig. 5]. Watching here is not necessarily used to fill “dead ing them on H6’s (in-use) broadcast-enabled TV; Rachel will
time” [39] but rather to spend downtime. For example, Fred default to on-demand watching on specific days of the week
(H6) has a particular YouTube routine on his phone: in order to watch from the comfort of her bed; and Tim will
“Usually when I get home from work I’ll sit down use specific device configurations in order to watch in HD.
for 10 minutes or so, about quarter of an hour, and
Media-multitasking
just sort of like right, before I do anything else,
have a choc ice [laughs], and go on YouTube for a The TV has been described “as a resource that can be dipped
bit.” (Fred, H6). into and out of as different activities come to dominate” [5,
pg. 15:14], leading to the accomplishment of other activities
The characteristic of portability alongside device quality can whilst the TV is on. More recently, “media-multitasking” [47]
lead users to borrow devices from others for watching; this has become familiar; many users access their mobile devices
is described by Kevin (H5) regarding his wife’s streaming, de- whilst watching TV for both “media-meshing” i.e. interac-
spite Ella owning a laptop herself: “Occasionally [Ella would] tion involving the TV program in view, and “media-stacking”
borrow my laptop to watch some Netflix thing whilst she’s i.e. interaction for other means unrelated to the TV [47].
cooking, just cause it had a bigger and better screen.”. Daily media-multitasking activities were common for 14/20
Other ways of watching can be more situational and spa- participants, consisting of: viewing social notifications (Ella,
tially determined. As the log data indicated that Rachel’s H5); checking emails (Ian, H7); shopping (Julie, H6); play-
(H8) Amazon Fire Stick demands more data on Tuesdays and ing games (Sally, H8); gathering news updates (Alan, H4);
Wednesdays, she described how going to bed early is the discovering information on TV shows (H5); smartphone use
cause of this. These days involve her using the TV dongle during TV adverts (Heather, H6); multi-watching (Fred, H6);
located in her bedroom (rather than her lounge TV) due to and other browsing (Laura, H2; H3; Denise, H4; Rachel, H8).
specific activities she routinely does or does not do: Five participants (i.e. Martin, Laura, Connie, Fred, Sally)
“Tuesday, erm I don’t see my boyfriend, he doesn’t mentioned that their media-multitasking occurs due to their
come round, so I go to bed when Sally goes to bed, disengagement with what they, or their household members,
and she comes in my bed for a bit and we watch are ‘watching’. Such multi-tasks can even cause conflict be-
telly, erm, and then Wednesday erm I go to my tween householders, as Ella said that “Kevin tells [her] off all
boyfriend’s for tea and then we come back here the time” when she checks notifications during TV watching.
and I don’t see the point in sitting in the front However, some participants do not media-multitask; despiteAlan (H4) having his iPad to hand while the TV is on, he go to sleep, I wanna watch the show”. Like Xavier, Ben is also
points out that he is not using them simultaneously: more content with Netflix than YouTube:
“I’m not doing both, I can’t multitask, so I might “I noticed when I had the Netflix subscription that
be concentrating perhaps on a newsfeed on the I was way more satisfied with the entertainment
iPad and the television’s on, or I’m watching the that I had, whereas YouTube entertainment is very
television and I’m just looking for an alternative like basic”...“it’s a bit like a slot machine isn’t it?
source of information or something.” (Alan, H4). Like yeah, it’s a bit of a gamble, like ‘you might
get something good, you might not’” (Ben, H1).
Furthermore, Ben (H1) stated that: “if I’m watching something
I’m watching”, ensuring focus on the content he is viewing. Ben reflected that services like YouTube are “designed to
Gemma (H1) even mentioned she has to have the TV turned hold your attention as long as possible”. As a result, he uses
off to “concentrate” on her online tasks: productivity tools and has edited a system file on his PC
to block access to sites he gets distracted by. Distractions
“I don’t even like the telly on when I’m, if I’m are not subject to YouTube alone, as Reddit can lead Kevin
using [my phone] or texting cause it’ll distract, (H5) to watch many short videos on the site. This ‘binge-
the telly will distract me from doing it”...“it’s a bit watching’ can occur with longer forms of video too, as Ella
like reading the same, when somebody’s talking to points out with her Netflix streaming: “I’ll be like ‘I’ll just
you and you’re reading the same page of a book, watch one more’”.
it just gets annoying doesn’t it?” (Gemma, H1).
As media-multitasking overlaps data demand from other 6 DISCUSSION
online activities, the demand linked to watching may be more Considering how technology is impacting watching norms,
than we present. With streaming becoming more prominent, we see a number of clear opportunities where HCI can chal-
the participants accounts of TV disengagement and the need lenge these new practices. We suggest we urgently need to
to focus on single online tasks, indicate that not all watching confront ‘all-you-can-eat’ and ‘binge’ watching more broadly,
demand will be fully absorbed or appreciated. as the shift to Internet based services has an increasing im-
pact on people, society and the planet. We point to the need
Trivial watching? for a broad framing to address political activism [37], radical
The participants have varying perspectives on what types societal transformation [36], and policy [67], to tackle this
of watching provides meaning to their lives, e.g. YouTube increasingly signficant concern. But we also acknowledge
is useful for learning to play instruments (Fred, H6; H7) or the need to evaluate SHCI designs with care [54] given the
for finding new musicians (Fred). However, general YouTube potential for unwanted rebound effects when addressing
viewing can be distracting: streaming in future work, as users could end up replacing
streaming with more energy-intensive activities; we note
“You can lose track of time sometimes”...“you go
our current discussion is limited in this regard.
on [YouTube] for something to do and you realise
you’ve been on there for half an hour because one The new norms
video of funny cats lead to another”...“sometimes
Shifts towards streaming and YouTube generation gaps. Our
it’s easier just to watch another video, ‘oh I’ll do
findings, coinciding with UK [50] and US [66] statistics, show
the ironing in a minute, oh there’s another video
that there have been significant shifts towards streaming as
of cats, I’ll watch that’” (Fred, H6).
a default. H5 no longer have a TV license. With Fred and
Despite Ben (H1) and Xavier (H9) regularly using YouTube, Alan, online content is emphasized as the primary medium
they seemed somewhat dissatisfied with the site’s content. for watching, with more traditional forms of viewing (i.e.
For example, Xavier falls asleep while watching YouTube broadcast TV, DVDs) becoming a secondary, if not obsolete,
videos he describes as “mundane” and as “background noise”— form of entertainment. Unsurprisingly the shift to streaming
the video content of which isn’t meaningful to him: is more prominent with younger generations (Generation Z,
“Most of it’s just junk content, people playing Millennials) accessing YouTube. Despite these shifts, the en-
games and then making jokes over the top of it”...“it’s ergy cost linked to these older infrastructures (e.g. broadcast
almost like a podcasty thing, I don’t really watch TV) is not necessarily removed; meaning streaming norms
it for the content itself but it’s more about the, the create another layer of energy impact.
voice overs...” (Xavier, H9). Watching as a distraction. Watching can become trivial: users
Xavier further explains how this YouTube watching contrasts can become disappointed with the time they spend watch-
to Netflix: “When I’m watching Netflix, I don’t really wanna ing (Ben, Fred, Kevin); and YouTube in particular, the mostcommon [55] and demanding watching service, can some- members. Co-creating these guidelines and technologies with
times only provide “mundane” or “distracting” entertainment. users could help them shift their streaming activities in more
Whilst paid streaming services can provide ‘better’ content meaningful, family-orientated and sustainable directions;
(Ben), they can also promote binge-watching (Ella). This moving away from watching “alone together” [69].
shows how different watching services and their content can
greatly change in meaning for users. Confronting ‘all-you-can-eat’ and ‘binge-watching’
‘All-you-can-eat’ contracts for home broadband and cellular
A screen (or two), each, anytime of the day. Our participants
data have become common, enabling multiple devices and
in shared houses illustrate how each person is becoming
media-rich interactions. Some contracts include unlimited
more focused on their own watching devices (H3, H6, H7,
data for streaming (e.g. Three’s “Go Binge” deal7 ) and bundle
H8, H9). This is even occurring whilst users are the same
media subscriptions with contracts (e.g. EE customers can get
room (H6)—an activity that Ofcom found for a third of UK
free BT Sport [19], Sprint customers can access Hulu [20]);
households [48]. The act of domestic multi-device watching,
this encourages media consumption, further propagating
i.e. multi-watching, is contributing to exaggerated evening
default streaming norms and growing data demand. ‘Binge-
peaks (figure 1) and overlaid demand (table 4). One of the
watchers’ and perhaps less disciplined consumers (e.g. Ella,
big challenges for HCI to overcome when considering how
Ben, Fred, Kevin) are aware that that they are captured
to tackle multi-watching is that interactivity can be core in
by auto-play [14] and the infinite availability of video on
these experiences. At a basic level, most online services aim
YouTube, Netflix and other forums e.g. Reddit.
to keep users engaged for longer, leading to more demand.
How is it that excess is valued as neutral or even positive, in
Reducing the data demand of watching through HCI this context? Taking health as an analogy (‘all-you-can-eat’
food, binge drinking), overload is seen negatively [1, 4, 38].
Limiting watching to the least data demanding configurations.
Yet, binge use drives major selling points of Internet con-
Interventions could steer users away from the new watching
tracts. Ultimately, if binge-ing is bad for our health, why are
possibilities that the multitude of household devices allow [6].
ISPs and service providers allowed to promote data gorging?
A greater shared understanding of devices and their capa-
There is a real need to rethink regulations (e.g. caps) on what
bilities may help constrain users to watch content in the
data demand or screen time providers can responsibly en-
least data demanding way. Take the example of Tim (H3)
courage, for the good of the user and the environment. Whilst
and his choices of either watching BT Sport in standard def-
previous HCI research has suggested that streaming services
inition on his YouView player, or in high definition on his
should help users gain more control over their watching
Google Chromecast via his smartphone: Tim chooses the lat-
sessions (e.g. by informing users when their ‘optimal’ view-
ter, yet devices and services could be designed to encourage
ing time has been reached or passed [14]), contributions to
the former (least data-intensive) device setup. Whilst previ-
over-watching are much more pervasive than the design of a
ously discussed ‘nudge’ approaches for opt-in high definition
particular app. Our findings point towards a need for a more
video [51] may encourage standard definition viewing for
responsible stance on the ‘all-you-can-eat’ philosophies and
some users, these are less likely to be effective for streaming
business models of ISPs and cellular providers—all of which
‘connoisseurs’ [6] like Tim. As a result, further encourage-
enable the prevalence of streaming in everyday life.
ment across devices will be required.
Co-creating what amount of streaming is ‘enough’. Our find- 7 IMPLICATIONS FOR HCI
ings show that household members are watching separately; Rethinking UX and Quality of Experience
and some streamed content can be trivial for users. The emer- Within HCI, media-multitasking is looked upon as a positive
gence of smart home devices could be used to combat the user experience [18, 24, 44]. With our participants, this mul-
associated demand: users’ current Internet activities could titasking means that streamed content is not always fully
be shared or made more visible (through displays, apps, or utilised (or “consumed”) and inherently makes watching
dis-aggregated views) in the household. Media-multitasking more data demanding. Here lies a tension: HCI promotes in-
could be omitted by prompting users to choose between novative and improved user experience (UX), whereas SHCI
watching and the secondary activity; and multi-watching highlights the need to be conscious of the utilisation and pro-
could be removed by bringing household members together motion of data demanding services [51, 72]. What if Quality
for group-only streaming [72]. Both of these examples could of Experience (QoE) considered reducing data demand?
exploit predictive algorithms to present options of what There is an opportunity to actively degrade QoE as a
households could do or watch at a particular point in time, HCI intervention, deterring the data demanding forms of
helping them to avoid any difficult activity choices or po-
tential conflicts in content preferences between household 7 http://www.three.co.uk/go-binge, accessed 31st August 2018.watching we present (e.g. multitasking and multi-watching) Developing a robust evidence base for policy makers
and helping users think about negotiable forms of watch- Internet policies driving “superfast” and “full fibre” access [26]
ing [7]. HCI could work more closely with network systems may only be fuelling more demand, as infrastructural capac-
researchers who are experts in, and drive the agenda of, ity growth leads to an increase in demand [51, fig. 1]. From
QoE [22, 23, 25]. Through this partnership, HCI researchers this standpoint, it is clear that policy makers have not made
and practitioners could: implement interfaces and services the connection between binge-watching and all-you-can-eat
that nudge and shift users towards less demanding modes marketing and data demand. Possibly blinded by the utility
of watching [51]; be more instrumental in promoting net- of Internet, there has been little discussion on the energy
work infrastructure running from renewable energy (e.g. impact of the Internet and its services within public policy
data centres [8]); and help create Internet standards that sup- (perhaps only a recent report by Policy Connect [42]). How
port sustainability (e.g. IETF8 , ISO9 ). Collaborations like this can HCI researchers help policy makers consider the growing
could highlight the data impact of interaction changes in environmental impacts associated with data demand?
digital services prior to implementation, preempting effects HCI researchers need to build robust knowledge bases
on network operators (e.g. Facebook’s effect by introducing and engage with creation of more responsible policy when it
video auto-play [59]). comes to ICT. Not only should new norms of everyday data
demand be considered (e.g. the new ways of watching we
Sustainable streaming contravenes net neutrality present), but also emerging Internet-based technologies (e.g.
One of the key points from our findings and discussion is cryptocurrencies, IoT, smart homes and connected cars) and
that we might limit traffic in different ways to create less data broader SHCI topics that require policy engagement [10, 16,
demand, with potentially profound impacts on society (e.g. 17, 67]. This will involve HCI researchers providing policy
impacting the revenue of content creators using YouTube makers with ideas (e.g. designs, interventions), and evidence
and Twitch). In particular, traffic limits are clearly at odds of their sustainable effect, that affect different aspects of HCI
with ideas of net neutrality; this takes all traffic as equal e.g. users themselves, interfaces HCI practitioners create, and
and at an equal cost, guaranteeing a fair level of access and subscription designs that service providers introduce. How
service to all. Our suggestions therefore coincide with the HCI researchers would then present these to policy makers
net neutrality repeal in the US [46] and conflict with policies is still in question, much like our net neutrality implication.
such as the EU’s Open Internet [13]. We are not against Perhaps the largest hurdle here in HCI (and at CHI) is finding
the social justice issues that EU policies protect; we suggest the appropriate venues to promote these discussions and
traffic limits for reasons of greater good (i.e. environmental affect change in systems, interfaces and policies.
sustainability) over increased profits for service providers.
As a result, if video traffic should cost more to reflect its cost 8 CONCLUSION
to the environment, these would have to be applied to all
In this paper, we have explored the most Internet-demanding
forms of video content; YouTube, Netflix, and the like would
activity in the home: watching. We have shown how video-
all have the same quota on watching traffic.
on-demand activities are composed, how these link to stream-
In some contexts, videos may still have to be treated differ-
ing services, and how they continue to evolve in new ‘data
ently (e.g. Emergency Broadcast System). This is a hard bal-
demanding’ ways. We have identified opportunities for the
ance to maintain and cannot be resolved by HCI researchers
HCI community to reduce watching data demand; highlight-
alone. Policy makers are also required to consider the so-
ing broader implications for HCI and society—showing how
cial, environmental and economic implications of surveilling,
HCI and other communities (e.g. policy makers, network en-
regulating and controlling portions of the Internet. HCI re-
gineers, service designers) do, and should, interlink. It is time
searchers should seek out policy makers for interdisciplinary
that we, as a society, work together to redefine our watching
investigations in this area. We are not the first researchers in
futures and begin dealing head-on with the unsustainable
SHCI to point out the need for wider political and societal in-
trajectory of this data demand.
volvement [10, 16, 17, 67]. To do this, HCI researchers could
begin by approaching governmental departments (e.g. UK
Department for Digital, Culture, Media, and Sport, US De- 9 ACKNOWLEDGEMENTS
partment of Commerce’s Digital Economy Agenda), or look We thank our participants for their cooperation. We also
to how previous HCI researchers have communicated with thank Matthew Broadbent for his support in this study. This
policy makers (e.g. International Policy Ideas Challenge [67]). work was supported by the Engineering and Physical Sci-
ences Research Council (grant number W95738G). Due to the
8 https://www.ietf.org/ small sample size, we have not made data publicly available
9 https://www.iso.org/ as it may compromise participant anonymity.REFERENCES conference on human factors in computing systems. ACM, 1975–1984.
[1] Tamás Agh, Gábor Kovács, Dylan Supina, Manjiri Pawaskar, Barry K [17] Paul Dourish. 2010. HCI and environmental sustainability: the politics
Herman, Zoltán Vokó, and David V Sheehan. 2016. A systematic of design and the design of politics. In Proceedings of the 8th ACM
review of the health-related quality of life and economic burdens of conference on designing interactive systems. ACM, 1–10.
anorexia nervosa, bulimia nervosa, and binge eating disorder. Eating [18] John Dowell and Edward Anstead. 2017. Interaction with a TV Com-
and Weight Disorders-Studies on Anorexia, Bulimia and Obesity 21, 3 panion App as Synopsis and Supplement. In Proceedings of the 2017 CHI
(2016), 353–364. Conference on Human Factors in Computing Systems. ACM, 2264–2268.
[2] Anders SG Andrae and Tomas Edler. 2015. On global electricity usage [19] EE. 2018. EE LAUNCHES ENHANCED BT SPORT APP OFFER
of communication technology: trends to 2030. Challenges 6, 1 (2015), WITH NEW TV CASTING AND INCLUSIVE MOBILE DATA. (2018).
117–157. http://newsroom.ee.co.uk/ee-launches-enhanced-b
[3] Joshua Aslan, Kieren Mayers, Jonathan G Koomey, and Chris France. t-sport-app-offer-with-new-tv-casting-and-inclu
2017. Electricity intensity of Internet data transmission: Untangling sive-mobile-data/, accessed August 2018.
the estimates. Journal of Industrial Ecology (2017). [20] USA TODAY Eli Blumenthal. 2017. Sprint’s latest bid to lure customers:
[4] Andrea M Barbarin, Laura R Saslow, Mark S Ackerman, and Tiffany C Free Hulu. (2017). https://eu.usatoday.com/story/tech
Veinot. 2018. Toward Health Information Technology that Supports /2017/11/15/sprints-latest-bid-lure-customers-fre
Overweight/Obese Women in Addressing Emotion-and Stress-Related e-hulu/863818001/, accessed August 2018.
Eating. In Proceedings of the 2018 CHI Conference on Human Factors in [21] Ericsson. 2018. Ericsson Mobility Report Interim Update. Technical
Computing Systems. ACM, 321. Report. Ericsson.
[5] Louise Barkhuus and Barry Brown. 2009. Unpacking the television: [22] Arsham Farshad, Panagiotis Georgopoulos, Matthew Broadbent, Mu
User practices around a changing technology. ACM Transactions on Mu, and Nicholas Race. 2015. Leveraging SDN to provide an in-network
Computer-Human Interaction (TOCHI) 16, 3 (2009), 15. QoE measurement framework. In Computer Communications Work-
[6] Oliver Bates, Mike Hazas, Adrian Friday, Janine Morley, and Adrian K shops (INFOCOM WKSHPS), 2015 IEEE Conference on. IEEE, 239–244.
Clear. 2014. Towards an holistic view of the energy and environmen- [23] Lyndon Fawcett, Mu Mu, Matthew Broadbent, Nicholas Hart, and
tal impacts of domestic media and IT. In Proceedings of the SIGCHI Nicholas Race. 2017. SDQ: Enabling rapid QoE experimentation us-
Conference on Human Factors in Computing Systems. ACM, 1173–1182. ing Software Defined Networking. In Integrated Network and Service
[7] Oliver Bates, Carolynne Lord, Bran Knowles, Adrian Friday, Adrian Management (IM), 2017 IFIP/IEEE Symposium on. IEEE, 656–659.
Clear, and Mike Hazas. 2015. Exploring (un) sustainable growth of [24] Tom Feltwell, Gavin Wood, Kiel Long, Phillip Brooker, Tom Schofield,
digital technologies in the home. (2015). Ioannis Petridis, Julie Barnett, John Vines, and Shaun Lawson. 2017.
[8] Lotfi Belkhir and Ahmed Elmeligi. 2018. Assessing ICT global emis- I’ve been manipulated!: Designing Second Screen Experiences for
sions footprint: Trends to 2040 & recommendations. Journal of Cleaner Critical Viewing of Reality TV. In Proceedings of the 2017 CHI Conference
Production 177 (2018), 448–463. on Human Factors in Computing Systems. ACM, 2252–2263.
[9] Eli Blevis. 2007. Sustainable interaction design: invention & disposal, [25] Panagiotis Georgopoulos, Yehia Elkhatib, Matthew Broadbent, Mu
renewal & reuse. In Proceedings of the SIGCHI conference on Human Mu, and Nicholas Race. 2013. Towards network-wide QoE fairness
factors in computing systems. ACM, 503–512. using openflow-assisted adaptive video streaming. In Proceedings of the
[10] Hronn Brynjarsdottir, Maria Håkansson, James Pierce, Eric Baumer, 2013 ACM SIGCOMM workshop on Future human-centric multimedia
Carl DiSalvo, and Phoebe Sengers. 2012. Sustainably unpersuaded: networking. ACM, 15–20.
how persuasion narrows our vision of sustainability. In Proceedings of [26] UK Government. 2017. Broadband Delivery UK. Technical Report.
the SIGCHI Conference on Human Factors in Computing Systems. ACM, Department for Culture, Media & Sport, UK Government. ht
947–956. tps://www.gov.uk/guidance/broadband-delivery-uk, ac-
[11] Meeyoung Cha, Haewoon Kwak, Pablo Rodriguez, Yong-Yeol Ahn, and cessed April 2018.
Sue Moon. 2007. I tube, you tube, everybody tubes: analyzing the [27] UK Government. 2017. Next Generation Mobile Technolo-
world’s largest user generated content video system. In Proceedings gies: A 5G Strategy for the UK. Technical Report. De-
of the 7th ACM SIGCOMM conference on Internet measurement. ACM, partment for Culture, Media & Sport, UK Government.
1–14. https://assets.publishing.service.gov.uk/governm
[12] Cisco. 2018. Cisco Visual Networking Index: Forecast and ent/uploads/system/uploads/attachment_data/file/
Trends, 2017–2022. Technical Report. Cisco. https: 597421/07.03.17_5G_strategy_-_for_publication.pdf,
//www.cisco.com/c/en/us/solutions/collateral/s accessed August 2018.
ervice-provider/visual-networking-index-vni/white [28] Lorenz M Hilty. 2015. Computing efficiency, sufficiency, and self-
-paper-c11-741490.pdf, accessed December 2018. sufficiency: A model for sustainability. In First Workshop on Computing
[13] European Commission. 2013, updated 2018. Open Internet. Technical within Limits (LIMITS 2015).
Report. European Commission. https://ec.europa.eu/digit [29] Alexis Hiniker, Hyewon Suh, Sabina Cao, and Julie A Kientz. 2016.
al-single-market/en/open-internet-net-neutrality, Screen time tantrums: how families manage screen media experiences
accessed August 2018. for toddlers and preschoolers. In Proceedings of the 2016 CHI Conference
[14] Dimph de Feijter, Vassilis-Javed Khan, and Marnix van Gisbergen. on Human Factors in Computing Systems. ACM, 648–660.
2016. Confessions of a’guilty’couch potato understanding and using [30] Christian Holz, Frank Bentley, Karen Church, and Mitesh Patel. 2015.
context to optimize binge-watching behavior. In Proceedings of the I’m just on my phone and they’re watching tv: Quantifying mobile
ACM International Conference on Interactive Experiences for TV and device use while watching television. In Proceedings of the ACM Inter-
Online Video. ACM, 59–67. national Conference on Interactive Experiences for TV and Online Video.
[15] Alex de Vries. 2018. Bitcoin’s Growing Energy Problem. Joule 2, 5 ACM, 93–102.
(2018), 801–805. [31] Yoori Hwang, HyoungJee Kim, and Se-Hoon Jeong. 2014. Why do
[16] Carl DiSalvo, Phoebe Sengers, and Hrönn Brynjarsdóttir. 2010. Map- media users multitask?: Motives for general, medium-specific, and
ping the landscape of sustainable HCI. In Proceedings of the SIGCHI content-specific types of multitasking. Computers in Human BehaviorYou can also read
